10
The impact of work accidents experience on causal attributions and worker behaviour So ´nia M. Pedroso Gonc ßalves a, * , Silvia Agostinho da Silva a , Maria Luı ´sa Lima a , Josep L. Melia ´ b a Centro de Investigac ßa ˜o e Intervenc ßa ˜o Social (CIS), Instituto Superior de Cie ˆncias do Trabalho e da Empresa (ISCTE), Lisboa, Portugal b Universitat de Vale `ncia, Vale `ncia, Spain Abstract It has recently been suggested that the experience of work accidents is an important variable to be considered as a pre- dictor of workers’ perceptions (e.g. causal attributions) and behaviours. Departing from the literature, this study has two goals: (1) to analyse the relationship among work accident experience, causal attribution of accidents and workers’ behav- iour; and (2) to test causal attributions as a mediating variable in the relationship between work accident experience and workers’ behaviour. To test the stability of the results, the same analyses have been performed in two Portuguese organi- zations, one in an industrial context and the other in an R&D context. In the industrial organization, the sample is com- posed of 559 employees and in the R&D organization the sample is composed of 335 employees. Results show that work accident experience is positively associated with external attributions and unsafe behaviours and negatively associated with internal attributions. Moreover, the results reveal a complete mediation of the causal attributions in the industrial orga- nization, although in the R&D organization the mediation was only partial. Ó 2007 Elsevier Ltd. All rights reserved. Keywords: Work accident experience; Causal attributions; Safety behaviours; Unsafe behaviours 1. Introduction Work accidents constitute an extremely serious problem in our society, given the important psychological, health, social, economical and organizational consequences associated with them (International Labour Orga- nization, 2003). This problem is reinforced by statistics, which reveal worrying numbers. Recent world data, from 2001 (International Labour Oce, 2005), indicates the occurrence of 268 million non-fatal and 351,500 fatal work accidents; in Europe the latest estimates, of the year 2003, allude to around 4.2 million work acci- dents resulting in more than 3 days of absence from work (EUROSTAT, 2005). 0925-7535/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.ssci.2007.11.002 * Corresponding author. Address: CIS/ISCTE, Edifı ´cio ISCTE, Av. das Forc ßas Armadas 1649-026 Lisboa. Tel.: +351217903079; fax: +351217903962. E-mail address: [email protected] (S.M.P. Gonc ßalves). Available online at www.sciencedirect.com Safety Science 46 (2008) 992–1001 www.elsevier.com/locate/ssci

The Impact of Work Accidents Experience on Causal Attributions and Worker Behaviour

Embed Size (px)

DESCRIPTION

Pedroso Et Al (2008)

Citation preview

  • The impact of work accidents experience on causalattributions and worker behaviour

    Sonia M. Pedroso Gonc!alves a,*, Silvia Agostinho da Silva a, Maria Lusa Lima a,Josep L. Melia b

    aCentro de Investigac!ao e Intervenc!ao Social (CIS), Instituto Superior de Ciencias do Trabalho e da Empresa (ISCTE), Lisboa, PortugalbUniversitat de Vale`ncia, Vale`ncia, Spain

    Abstract

    It has recently been suggested that the experience of work accidents is an important variable to be considered as a pre-dictor of workers perceptions (e.g. causal attributions) and behaviours. Departing from the literature, this study has twogoals: (1) to analyse the relationship among work accident experience, causal attribution of accidents and workers behav-iour; and (2) to test causal attributions as a mediating variable in the relationship between work accident experience andworkers behaviour. To test the stability of the results, the same analyses have been performed in two Portuguese organi-zations, one in an industrial context and the other in an R&D context. In the industrial organization, the sample is com-posed of 559 employees and in the R&D organization the sample is composed of 335 employees. Results show that workaccident experience is positively associated with external attributions and unsafe behaviours and negatively associated withinternal attributions. Moreover, the results reveal a complete mediation of the causal attributions in the industrial orga-nization, although in the R&D organization the mediation was only partial.! 2007 Elsevier Ltd. All rights reserved.

    Keywords: Work accident experience; Causal attributions; Safety behaviours; Unsafe behaviours

    1. Introduction

    Work accidents constitute an extremely serious problem in our society, given the important psychological,health, social, economical and organizational consequences associated with them (International Labour Orga-nization, 2003). This problem is reinforced by statistics, which reveal worrying numbers. Recent world data,from 2001 (International Labour Office, 2005), indicates the occurrence of 268 million non-fatal and 351,500fatal work accidents; in Europe the latest estimates, of the year 2003, allude to around 4.2 million work acci-dents resulting in more than 3 days of absence from work (EUROSTAT, 2005).

    0925-7535/$ - see front matter ! 2007 Elsevier Ltd. All rights reserved.doi:10.1016/j.ssci.2007.11.002

    * Corresponding author. Address: CIS/ISCTE, Edifcio ISCTE, Av. das Forc!as Armadas 1649-026 Lisboa. Tel.: +351217903079; fax:+351217903962.

    E-mail address: [email protected] (S.M.P. Gonc!alves).

    Available online at www.sciencedirect.com

    Safety Science 46 (2008) 9921001

    www.elsevier.com/locate/ssci

  • The last few decades have been marked by several multidisciplinary contributions to the prevention of workaccidents, mainly focusing on those factors that predict workers behaviours in relation to prevention and risk.Scientific studies have focused their attention, primarily, on the role that human and management factors playin safety (Hale and Hovden, 1998). The literature has suggested that several variables at an organizational andgroup level, such as the organizational and group safety environment (e.g. Neal et al., 2000; Gonc!alves et al.,2005, and at an individual level, cognitions such as risk perception, causal attribution, as well as personal acci-dent experience (Cree and Kelloway, 1997; Melia`, 1998; Lima, 1999), have played an important role in pre-dicting workers behaviours. Nevertheless, the analysis of the literature shows that the results of the studiesare inconclusive, and even incongruent in some situations.

    The present study was developed with the principal concern of contributing to a better understanding of thevariables that predict workers behaviours in relation to safety, namely, to explore the role of personal acci-dent experience and causal attributions.

    1.1. Causal attributions

    Many years ago, Heider (1958) called our attention to the fact that each of us, in the process of makingsense of our physical and social world, acts as a naive psychologist, trying to explain our actions and theactions of others. By casual attributions he referred to the tendency of people to offer explanations by assess-ing the logical association between the cause and effect variables (Heider, 1958). The major contribution ofHeider was his crucial distinction between personal/internal and situational/external causes (Leyens and Yzer-byt, 1999), as two types of explanations for an event with very different consequences. For example, work acci-dents being explained as a result of the inappropriate behaviour or stupidity of a worker (internal explanation)or as the consequence of a problem with machinery or noise in the work environment (external explanation)lead to different approaches towards safety.

    The literature on attribution (e.g. Nisbett and Ross, 1980; Jones, 1990; Melia` et al., 2001) suggests thatunexpected negative events generate a higher number of attributions. Work accidents fulfil this criterion,and thus stimulate the search for a causal explanation (Melia` et al., 2001). According to Kouabenan(2002), research on risk in general, and accidents in particular, elicits an intensive cognitive activity in peopleaimed at finding reassuring explanations and gaining a better sense of control over the situation. This is one ofthe functions of causal attributions, and one of the reasons why this process involves some systematic bias.Some authors have described defensive attributions in the explanation of personal experience of accidentslinked with the fact that subjects, to protect their role in the situation, minimize their own responsibilityand maintain their self-esteem, tending to make more external causal attributions (e.g. Hewstone, 1989; Salmi-nen, 1992; Kouabenan et al., 2001); what has been called the fundamental attribution error, refers to the factthat observers tend to attribute events to the individuals involved in them, rather than to situational factors(e.g. Hewstone and Antaki, 1990; Melia` et al., 2001). Moreover, according to Weiner (1985), the type of causalattribution produced is connected to the expectations that influence future performance. Internal attributionsof failures, especially when they are also unstable and controllable (such as the attribution of an accident tolack of care or lack of effort) are associated with a more persistent behaviour in future occasions; whileexternal attributions (such as bad luck or influence of others) and internal stable (such as I am a careless per-son) will lead to a lack of motivation to improve behaviour as it is seen as out of personal control. Accordingto Gykye (2003, p. 533), understanding the underlying causes of accidents provides knowledge that guidesthe behaviour of workers. Nevertheless, no study has explored how causal attributions of work accidentsinfluence workers behaviours.

    1.2. Experience of work accidents

    Since the beginning of the twentieth century, the accident proneness theory has suggested that work acci-dent experience influences workers cognitions and behaviours. However, only a small number of studies haveexplored how this variable influences workers perceptions and behaviours (Cree and Kelloway, 1997; Gold-berg et al., 1991). For instance, in some studies, positive correlations were found between accident experience

    S.M.P. Gonc!alves et al. / Safety Science 46 (2008) 9921001 993

    Eduardo Aguirre Dvila

  • and safety behaviour, suggesting that accident experience appears to result in more cautionary behaviour (e.g.Laughery and Vaubel, 1989; Kouabenan, 2002).

    These results, however, should be understood using models that explain why and when accident experiencegives rise to safety behaviours, since the cognitive approach suggests that accident experience influences work-ers evaluation and perception, which in turn influences peoples behaviour (Laughery and Vaubel, 1989).

    Causal attributions can be one of those mediating variables, but studies that explore the relationshipbetween experience of work accidents and causal attributions of work accidents are even less frequent andthe results are not clear. For example, Kouabenan (2002), contrary to his predictions, found that accidentexplanations were not affected by whether or not the person had been the victim of an accident in the past.

    There are a small number of studies that have explored the personal work accident experience and until thepresent work no study has focused simultaneously on workers unsafe and safe behaviours. Safety behaviourrefers to all safety performance in the workplace, which affects the probability of accidents, unhealthy andother undesired results. Although the quality and level of safety behaviour at work is complex and multidi-mensional, the most relevant safety behaviours can be identified as safe behaviours, i.e. those behaviours thatcontribute to reduce the probability of accidents, and unsafe behaviours, i.e. those involved in the increasingof the risk or probability of accidents (Melia, 2007). There are some safe behaviours opposite and incompat-ible (e.g. use and not use a personal protective device), but many times employees perform some safe andunsafe behaviours simultaneously, or nearly simultaneous. Our observations have taught us that unsafebehaviours and safe behaviours are not necessarily incompatible with each other and they can co-exist. Forinstance, it is possible to see a worker using his/her helmet and gloves (safe behaviour) and breaking safetyrules to get the work done faster (unsafe behaviour).

    1.3. The present study

    This study intends to contribute towards the clarification of the association between work accident expe-rience and behaviours from a cognitive approach. Hence, departing from this literature and considering thathaving experienced an accident will have an impact on individual behaviour, one theoretical model was tested.The model hypothesizes a mediation effect, suggesting that the influence of workers accident experience influ-ences causal attributions, which in turn influences workers behaviours (see Fig. 1). Moreover, this modelincludes both external and internal attributions, as well as safe behaviour and unsafe behaviour.

    From this general model, some hypotheses were formulated. Work accidents are expected to be positivelyassociated with external attributions and negatively associated with internal attributions. External attributionsare also expected to be positively associated with unsafe behaviours, while internal attributions are expected tohave a negative association, and the inverse correlation pattern is expected for the relation between attribu-tions and safe behaviours.

    2. Method

    2.1. Sample

    In the present study, the model has been tested in two samples from two different Portuguese organizationswith different contexts in the safety field: an organization from the industrial sector composed of 559 parti-cipants and an organization from the research and development sector composed of 335 participants. Thesample stratification was based on the relative size of the various departments in the companies, resulting

    Work Accident Experience

    CausalAttributions

    - External- Internal

    WorkerBehaviour

    - Safety- Risk

    Fig. 1. Model for worker behaviour and causal attribution.

    994 S.M.P. Gonc!alves et al. / Safety Science 46 (2008) 9921001

  • in a random sample in which the various departments were represented proportionally. In previous studies inthese two organizations, it was possible to conclude that both are characterized by a low concern with safety.

    In the first organization, an industrial company (IO), all the participants were men, and the majority(60.0%) had ages between 31 and 50 years old. The majority (46.7%) of these participants had worked inthe company for more than 20 years, at all hierarchical levels, although the majority (76.3%) were operators.

    In the second organization (R&D), a research and development organization from the services sector, themajority of the participants were women (67.2%), with ages between 31 and 50 years old (53.3%) and withmore than 17 years of seniority in the organization (73.3%). An important characteristic of this organizationwas that it did not have a safety department or a safety supervisor, and neither did it have any formal safetynorms.

    2.2. Measures

    In the two studies, all the variables were evaluated on the basis of a self-report questionnaire, using similarscales.

    2.2.1. Accident experienceThis was measured by three questions that evaluated self-report of work accidents (e.g. How many acci-

    dents have you had in this company?, How many incidents have you had in the last 5 years?, How manyaccidents with absence from work have you had in the last 5 years?). The responses to the three items weremerged to create four levels of gravity of work accident experience for each employee (the lowest level of workaccident experience: have never had accidents in this company, neither incidents or accidents in the last 5years; the highest level of work accident experience: have had accidents in this company, incidents and severeaccidents in the last 5 years).

    2.2.2. Causal attributionsThe causal attributions were assessed by a method similar to that used in previous attributions studies (e.g.

    Hofman and Stetzer, 1998; Kouabenan et al., 2001) with two scales, the external attributions scale focused onthe organizational environment (six items; e.g. The majority of work accidents that occurred in my companyare related with the lack of inspection from the governmental authorities; a10 = 0.76; aR&D = 0.75) and theinternal attributions scale focused on the workers (three items; e.g. The majority of work accidents that haveoccurred in my company are related with the non-fulfilment of the rules of safety by the workers; aIO = 0.77;aR&D = 0.76). These scales were confirmed in a principal component factor analysis. In each case, responseswere given along a 7-point scale (1 = strongly disagree, 7 = strongly agree), in which participants had to iden-tify the causes of work accidents in their companies.

    2.2.3. Workers behavioursWorkers behaviours were measured on the basis of two scales: safe behaviours (adapted from Neal and

    Griffin, 2004) and unsafe behaviours (based on Rundmo et al., 1998). The division between safe and unsafebehaviour was confirmed through a principal component factor analysis in both organizations. The safebehaviours scale includes four items (e.g. I communicated the existence of dangerous situations in my work-place) with acceptable reliability levels (aIO = 0.74; aR&D = 0.46; In the last sample, the alpha value is low,but the small number of items can partially justify this result. For this reason, homogeneity tests were per-formed, following Brigg and Cheek, 1986 recommendations. Inter-item correlations range from 0.17 to0.25, with a mean value of 0.22, which is an acceptable value for an applied study). Unsafe behaviour wasmeasured with two items (e.g. I take risks to do the work faster; rIO = 0.25**; rR&D = 0.31**). In each case,responses were given along a 7-point scale (1 = never, 7 = always).

    3. Results

    The results were obtained in three phases: descriptive statistics, correlation estimation and regression esti-mation to test the model.

    S.M.P. Gonc!alves et al. / Safety Science 46 (2008) 9921001 995

  • 3.1. Descriptive statistics

    The results obtained (see Table 1) reveal that the industrial organization is characterized by a low level ofexperience of workplace accidents, higher levels of safe behaviours and low levels of unsafe behaviours. Incontrast, the results obtained reveal that the research and development organization is characterized by alow level of experience of workplace accidents, lower levels of safe behaviours and unsafe behaviours. Theresults show consistently that the participants make more internal attributions than external attributions toexplain the accidents that occurred in their organizations. Furthermore, it is possible to observe higher valuesof safe behaviours than unsafe behaviours.

    3.2. Correlations

    The correlation matrix (see Table 2) in the industrial organization shows that work accident experience hasa positive correlation with external causal attributions (r = 0.247, p = 0.000, N = 396), unsafe behaviours(r = 0.157, p = 0.001, N = 419) and safe behaviours (r = 0.106, p = 0.031, N = 415) and a negative correlationwith internal attributions (r = !0.112, p = 0.022, N = 414). Hence, the results reveal that the experience ofwork accidents is significantly associated with external attributions (and less with internal attributions) andboth with more unsafe and safe behaviours. In the research and development organization the correlationmatrix (see Table 2) reveals that work accident experience has a positive correlation with external causal attri-butions (r = 0.171, p = 0.011, N = 223) and unsafe behaviours (r = 0.244, p = 0.000, N = 237) and a negativecorrelation with internal attributions (r = !0.232, p = 0.000, N = 230). Considering the hypothesis stated, apositive correlation was found in the two companies for the relationship between unsafe behaviour and exter-nal attributions, as predicted, but the negative association with internal attributions was only significant forthe Industrial Company. The pattern of results for safe behaviour is not the reverse for unsafe behaviour, andthe correlations are non-significant.

    3.3. The test of the model mediation effect

    The model was tested following the recommended mediation procedures proposed by Baron and Kenny(1986), and reinforced recently by Frazier et al. (2004). According to this method there are four steps to

    Table 2Correlations among the variables in the two companies

    Industrial organization Research and development organization

    1 2 3 4 1 2 3 4

    1. Work accident experience 2. External attribution 0.247** 0.171*

    3. Internal attribution !0.112* 0.016 !0.232** !0.0534. Safe behaviour 0.106* 0.171** !0.061 0.118 0.006 0.0855. Unsafe behaviour 0.157** 0.289** !0.126** !0.087 0.244** 0.330** !0.023 !0.015Note. *p < 0.05; **p < 0.01.

    Table 1Descriptive statistics and reliability values for the data collected in the two companies

    Industrial organization Research and development organization

    Variables M SD a M SD a

    Work accident experience 2.08 1.08 1.68 0.99 External attribution 4.03 1.30 0.76 5.08 1.41 0.75Internal attribution 3.86 1.22 0.77 3.63 1.55 0.76Safe behaviour 5.14 1.16 0.74 4.37 1.61 0.78Unsafe behaviour 3.49 1.31 R = 0.25 3.05 1.63 R = 0.31

    Note. M (mean); SD (standard deviation); a (Cronbachs alpha).

    996 S.M.P. Gonc!alves et al. / Safety Science 46 (2008) 9921001

  • establish that a variable mediates the relation between a predictor variable and an outcome variable. The firststep is to show that there is a significant relation between the predictor and the outcome (it is possible to seethis in the correlation matrix). The second step is to show that the predictor is related to the mediator. Thethird step is to show that the mediator is related to the outcome variable. The final step is to show thatthe strength of the relation between the predictor and the outcome is significantly reduced when the mediatoris added to the model. If the mediator variable is a complete mediator, the relation between the predictorvariable and the outcome variable will not differ from zero after the mediator variable is included in the model.If the mediator variable is a partial mediator, the relation between the predictor and the outcome will besignificantly smaller when the mediator variable is included, but will still be greater than zero. In all stepsthe demographic variables were controlled (tenure, work, occupational category and age), which means thatthese variables are entered in the regressions, but due to reasons of space are not represented in the mediationstables presented subsequently.

    Considering the previous correlation matrix (step 1), it was possible to observe that in the research anddevelopment organization the results show some non-significant associations, which do not allow us to testthe performance of the model. Specifically, safe behaviour was not statistically associated with any of the vari-ables in the study. It is thus not possible to test the model in terms of safe behaviour prediction. The corre-lation between unsafe behaviours and internal attributions is not statistically significant, so it is not possible totest the model that hypothesizes the mediation effects of these types of attributions, for this organization. Inthe industrial organization it is not possible to test the mediator effect of internal attributions.

    3.3.1. Mediation effect of external attributionsExternal attribution was the first mediator tested.In the second step the effects of the independent variable (work accident experience) on the mediator was

    tested for both organizations. In the industrial organization the results reveal that work accident experienceexplains 14% of external attributions variance (F(5,244) = 9.309, p = 0.000; b = 0.190, p = 0.003). In theresearch and development organization it explains 2% of external attributions variance (F(6,166) = 1.693,p = 0.126; b = 0.188, p = 0.015).

    In the third and fourth steps the effect of work accident experience and external attributions in the predic-tion of unsafe and safe behaviour was tested (see Table 3). In the industrial organization, it was not possible toconfirm the mediation model for safe behaviours. Nevertheless, results for unsafe behaviour revealed thatexternal attributions are a complete mediator, since the relation between the predictor variable (work accidentexperience) and the outcome variable (unsafe behaviour) is no longer significant with the introduction of themediator variable in the model (external attributions). The proposed model was then supported only forunsafe behaviours. This result confirms that work accident personal experience has an impact on workersunsafe behaviour through external causal attributions.

    In the research and development organization, only the prediction of unsafe behaviour was possible. Exter-nal attributions are a partial mediator, since the relation between the predictor variable (work accident expe-rience) and the outcome variable (unsafe behaviour) became more significant with the introduction of the

    Table 3Regression for predicting safe and unsafe behaviour: external causal attributions as mediator

    Industrial organization Research and development organization

    Safe behaviour Unsafe behaviour Unsafe behaviour

    b DR2 FChange df b DR2 FChange df b DR

    2 FChange df

    1. WA 0.120 0.013 3.081 1241 0.186** 0.030 8.349 1244 0.271* 0.071* 13.245 1162R2 adjusted = 0.002 R2 adjusted = 0.095** R2 adjusted = 0.105*

    2. WA 0.096 0.118 0.211**

    EA 0.124 0.013 3.228 1240 0.362* 0.110 34.400 1243 0.313* 0.092* 19.204 1161R2 adjusted = 011 R2 adjusted = 0.204* R2 adjusted = 0.185*

    Note. *p < 0.001; **p < 0.05; ***p < 0.01: WA (work accident experience); EA (external attributions).

    S.M.P. Gonc!alves et al. / Safety Science 46 (2008) 9921001 997

  • mediator variable in the model (external attributions). However, the b value is smaller with the inclusion of themediator variable.

    3.3.2. Mediation effect of internal attributionsThe second mediation effect that was analysed was the mediation effect of internal attributions.In the second step the effects of the independent variable (work accident experience) on the mediator (inter-

    nal attributions) for both organizations were tested. The results, for the industrial organization, reveal thataccident experience explains 6% of internal attributions (F(5,253) = 4.436, p = 0.001; b = !0.097,p = 0.131), while in the research and development organization it also explains 6% of internal attributions var-iance (F(6,172) = 3.024, p = 0.008; b = !0.183, p = 0.014).

    In the next step, the effect of work accident experience and the mediating role of internal attributions in theprediction of unsafe and safe behaviours were tested (see Table 4).

    In the industrial organization, since the internal attributions do not seem to have a significant associationwith safe behaviour, the prediction of this type of behaviour was not performed. The results obtained for theprediction of unsafe behaviour, in the industrial organization, revealed that the mediation model was not sup-ported, as there is no relationship between internal attributions, the predictor variable (work accident experi-ence) and the outcome variable (unsafe behaviour).

    In the research and development organization it was not possible to test the mediation model, since theassociation between the variables was not statistically significant.

    4. Discussion of results

    The current paper presents a study that aims to contribute to the understanding of work accident impact onfuture behaviour and the role of causal attributions. This study is distinguished from previous research due tothe fact that it looks for behavioural consequences of work accident experience and causal attributions asso-ciated with work accidents. Trying to understand workers behaviours and identify their determinants is one ofthe first steps to prevention.

    This study had two objectives: (1) to analyze the relationship among work accident experience, causal attri-bution and workers behaviours; and (2) to test causal attributions as a mediating variable in the relationshipbetween work accident experience and workers behaviours. To test these relationships between variables astudy was conducted in two organizations.

    The results consistently show, for both organizations, that work accident experience has a positive corre-lation with external causal attributions and unsafe behaviour. It is thus possible to say that greater work acci-dent experience and higher external attributions related with organizations fault for accidents are associatedwith higher unsafe behaviour. These results are consistent with Weiner (1985) predictions based on his moti-vational model; and following Kouabenan (2002) predictions, the results consistently show that work accidentexperience influences causal attributions. Similar to the literature (e.g. Salminen, 1992; Kouabenan et al.,2001) the participants of the present study tends to make more external causal attributions than internal attri-butions; this tendency may be explained as a defensive mechanism minimizing workers responsibility.

    Table 4Regression for predicting unsafe behaviour: internal causal attributions as mediator

    Industrial organization, unsafe behaviour

    b DR2 FChange df

    1. WA 0.167** 0.025** 6.923 1252R2 adjusted = 0.089**

    2. WA 0.164**

    IA !0.029 0.001 0.223 1251R2 adjusted = 0.087

    Note. *p < 0.001; **p < 0.05; * **p < 0.01: WA (work accident experience); IA (internal attributions).

    998 S.M.P. Gonc!alves et al. / Safety Science 46 (2008) 9921001

  • In the industrial organization the results also reveal a negative association between unsafe behaviour andinternal attributions. However, in the research and development organization this association was not statis-tically significant. In relation to safe behaviour, although results revealed a positive association with workaccident experience and external attributions, the regression results did not confirm their predictive value.

    Accident experience will probably change workers behaviour (both safe and unsafe behaviour), at leastduring a certain period after the accident. This means that direct effects of accident experience and accidentseverity could be detected if safe behaviours were registered at the appropriate time, but that would implya longitudinal study which was not done in this case.

    From a cognitive approach, it is reasonable to expect that accident experience does not only affect employ-ees behaviour, but also the interpretation that workers make about accidents, which consequently also affectsworkers behaviours. Accordingly, our results (Fig. 2) consistently showed that external causal attributionsmediated the relationship between work accident experience and unsafe behaviour. The other associationswere non-significant. This complex set of relationships indicates that the specific characteristics of the organi-zational setting (for example, the organizations activity sector, the safety and preventive structures of theorganization or the demographic composition of the samples) may play an important role in this process.Moreover, in this study, as both companies are characterized by a low level of experience of workplace acci-dents, few causal attributions or other effects can be envisaged. Besides, the R&D company also has low levelsof safe behaviour and unsafe behaviour so, definitively in the R&D organization safety is not a salient issueand this makes it more difficult to observe causal attributions and their effects.

    In addition, causal attribution related to work accidents is not a spontaneous process. It is also a matter oftraining and information. A well-informed employee will rarely attribute accidents to only one cause. Baraoet al. (2006) revealed that workers belonging to companies with stronger positive safety cultures (e.g. with verygood safety training and communication) interpreted and explained work accident causes with a more com-plex approach, combining internal and external attributions. Therefore, causal attributions are not only a mat-ter of an individual perception bias, they are influenced by the organizational context and the role safety playsin this context. Safety culture and safety training will determine accident attributions and perhaps, the greaterthe safety training of the employees the more accurate their causal attributions of an accident will be.

    It is also important to stress that accident experiences may be very diverse quantitatively and qualitatively.In this study, the measure used tried to merge quality and quantity, but it focused more on the quantitativeaccumulation of accident experiences; perhaps the quality of the experience (i.e. the severity of the accident orthe severity of the social consequences, i.e. court processes) has more impact than the quantity. According toliterature review (e.g. Kouabenan, 1999), it is possible that the greater the severity of the accident the morecausal attributions are made.

    In conclusion, results also suggest that safe behaviour and unsafe behaviour have different determinants;and it seems to be easier to predict unsafe behaviour than safe behaviour. Work accident experience seemsto be a good predictor of unsafe behaviour, and this relationship is mediated by workers external attributions.However, the reverse pattern was not found for safe behaviour. Moreover, although significant relationshipsin five of the six regression equations for predicting safe and unsafe behaviours were found, the percentage ofvariance explained by predictors was low, which suggests that there is still a long walk in the research in con-nection with predicting workers behaviours.

    External Causal Attributions

    St2: IO = .190**

    St1: IO = .186**

    St3: IO = .362*

    St3: IO = .118St1: R&D = .271* St3: R&D = .211*

    St2: R&D = .188** St3: R&D = .313*Worker Unsafe Behaviour Work Accident Experience

    Fig. 2. Synthesis of the mediation effects (the steps of the analysis are indicated as St1St3). Note. Results above and below the linescorrespond to the Industrial Organization and the R&D company, respectively. *p = 0.001; **p = 0.05.

    S.M.P. Gonc!alves et al. / Safety Science 46 (2008) 9921001 999

  • 5. Conclusion

    Considering the knowledge accumulated in the literature, the results of these studies have some implicationsfor changing organizational behaviour. Work accident experience performs a central role affecting workersperceptions and behaviours, suggesting that organizational learning from accidents can be important at theindividual level, especially if the attribution pattern is considered. This study enhances the importance offocusing on the attribution of accidents in the workplace after an accident, and it suggests that unsafe andsafe behaviours should be considered as somewhat different realities.

    The generalization of the conclusions of these studies should be contextualized, taking into considerationsome limitations of the current study. This is a cross-sectional study based on self-reported data, which meansthat causality cannot be inferred and data may be biased. Some of the measures used should also be refined.Moreover, the study focused on the observers point of view, but the participants had to explain the accidentsthat occurred in their organization and not their own accidents. Some other contextual variables could alsohave been useful to contextualize our results, namely, the organizational and group safe culture.

    In future, research should continue to focus on the role of accident experience and its influence on howpeople make sense of work accidents to increase its positive impact for safe and unsafe behaviour. A zero acci-dents goal is a difficult goal to accomplish but the study of accidents causes is essential for prevention. Learn-ing from accidents and incidents is a social responsibility of all companies.

    Acknowledgements

    This research was conducted by the Centro de Investigac!ao e Intervenc!ao Social (CIS) Portugal and sup-ported by the FCT-Portugal (PIQS/PSI/50070/2003).

    Part of this paper was published earlier in Gonc!alves, S., Silva, S., Melia`, J., Lima, L., 2006. The experienceof work accidents: its consequences for cognition and behaviour. In: Guedes Soares, C., Zio, E., (Eds)., Safetyand Reliability for Managing Risk. London: Taylor & Francis Group, pp. 333339.

    References

    Barao, S., Silva, S.A., Lima, M.L., 2006. The role of safety culture in explaining work accidents. In: Guedes Soares, C., Zio, E. (Eds.),Safety and Reliability for Managing Risk. Taylor & Francis Group, London, pp. 751757.

    Baron, R.M., Kenny, D.A., 1986. The moderatormediator variable distinction in social psychological research: conceptual, strategic andstatistical considerations. Journal of Personality and Social Psychology 51, 11731182.

    Brigg, S.R., Cheek, J.M., 1986. The role of factor analysis in the development and evaluation of personality scales. Journal of Personality54 (1), 106148.

    Cree, T., Kelloway, K., 1997. Responses to occupational hazards: exit and participation. Journal of Occupational Health Psychology 2 (4),304311.

    Eurostat, 2005. The European Statistics on Accidents at Work: Initial results. http://www.av.se/dokument/statistik/english/Accidents2003_EU15.pdf.

    Frazier, P., Baron, K., Tix, A., 2004. Testing moderator and medator effects in counseling psychology research. Journal of CounselingPsychology 51 (1), 115134.

    Goldberg, A., Dar-El, E., Rubin, A., 1991. Threat perception and the readiness to participate in safety programs. Journal ofOrganizational Behavior 12, 109122.

    Gonc!alves, S., Silva, S., Melia`, J., Lima, M.L., 2005. Safety climate, risk perception and safety behavior [Clima de seguranc!a, percepc!ao derisco e comportamentos de seguranc!a]. In: Guedes Soares, C., Teixeira, A.P., Antao, P. (Eds.), Analysis and Management of Risks,Safety and Reliability [Analise e Gestao de Riscos, Seguranc!a e Fiabilidade]. Edic!oes Salamandra, Lisboa, pp. 119132.

    Gyekye, S.A., 2003. Causal attributions of Ghanian workers for accident occurrence: Miners and non-miners perspective. Journal ofSafety Research 34, 533538.

    Hale, A.R., Hovden, J., 1998. Management and culture: the third age of safety. A review of approaches to organizational aspects of safety,health and environment. In: Feyer, A.M., Williamson, A. (Eds.), Occupational Injury: Risk, Prevention and Intervention. Taylor &Francis, London, pp. 129165.

    Heider, F., 1958. The Psychology of Interpersonal Relations. Wiley, New York.Hewstone, M., Antaki, C., 1990. Theory of social attributions and explications [Teora de la atribucion y explicaciones sociales]. In:

    Hewstone, M., Strobe, W., Codol, J.P., Stephenson, G.M. (Eds.), Introduction to Social Psychology (Introduccion a la PsicologaSocial). Areil Psicologa, S.A, Barcelona, pp. 120148.

    Hewstone, M., 1989. Causal Attribution: From Cognitive Processes to Collective Beliefs. Basil Blackwell, Oxford.

    1000 S.M.P. Gonc!alves et al. / Safety Science 46 (2008) 9921001

  • Hofman, D.A., Stetzer, A., 1998. The role of safety climate and communication in accident interpretation: Implications for learning fromnegative events. Academy of Management Journal 41 (6), 644657.

    International Labour Organization, 2003. Safework: Accident and disease information. http://www.ilo.org/public/english/protection/safework/accidis/index.htm.

    International Labour Office, 2005. World Day for Safety and Health at Work 2005: A Background Paper. http://www.ilo.org/public/english/bureau/inf/download/sh_background.pdf.

    Jones, E., 1990. Interpersonal Perception. New York.Kouabenan, D., 1999. Explication nave de l accident et prevention. PUF, Paris.Kouabenan, D., 2002. Occupation, driving experience, and risk and accident perception. Journal of Risk Research 5 (1), 4968.Kouabenan, D., Gilbert, D., Medina, M., Bouzon, F., 2001. Hierarquical position, gender, accident severity, and causal attribution.

    Journal of Applied Social Psychology 31 (3), 553575.Laughery, K.R., Vaubel, K.P., 1989. The role of accident experiences on subsequent accident events. In: Feyer, A.M., Williamson, A.

    (Eds.), Occupational Injury: Risk, Prevention and Intervention. Taylor & Francis, London, pp. 3343.Leyens, J.-Ph., Yzerbyt, V., 1999. Social PSychology, Edic!oes 70, Lisboa, Psicologia Social.Lima, M., 1999. Risk perception and safety cultures in organizations. Psychology (Psicologia) XII (2), 379386, Percepc!ao de riscos e

    culturas de seguranc!a nas organizac!oes.Melia, J.L., 2007. The Human Factor in Occupational Safety. Occupational Safety and Health Psychology. Lettera Publicaciones, Bilbao,

    El Factor Humano en la Seguridad Laboral. Psicologia de la Seguridad y Salud Laboral.Melia`, J.L., Chisvert, M., Pardo, E., 2001. Procedural model of the attributions and attitudes towards work accidents: Measurement and

    intervention strategies. Journal of Work and Organizational Psychology (Revista de Psicologa del Trabajo y de las Organizaciones) 17(1), 6390, Un Modelo Processual de las atribuciones y Actitudes ante los Accidents de Trabajo: Estrategias de medicion eintervencion.

    Melia`, J.L., 1998. A psychosocial causal model of work accidents. Psychology Yearbook (Anuario de Psicologa) 29 (3), 2543, Un modelocausal psicosocial de los accidentes laborales.

    Neal, A., Griffin, M.A., 2004. Safety climate and safety at work. In: Barling, J., Frone, M.R. (Eds.), The Psychology of Workplace Safety.APA, Washington, pp. 1534.

    Neal, A., Griffin, M.A., Hart, P.M., 2000. The impact of organizational climate on safety climate and individual behaviour. Safety Science34, 99109.

    Nisbett, R., Ross, L., 1980. Human Inference: Strategies and Shortcomings of Social Judgment. Prentice-Hall, New Jersey.Rundmo, T., Hestad, H., Ulleber, P., 1998. Organisational factors, safety attitudes and workload among offshore oil personnel. Safety

    Science 29, 7587.Salminen, S., 1992. Defensive attribution hypothesis and serious occupational accidents. Psychological Reports 70, 11951199.Weiner, B., 1985. An attributional theory of achievement motivation and emotion. Psychological Review 92 (4), 548573.

    S.M.P. Gonc!alves et al. / Safety Science 46 (2008) 9921001 1001

    The impact of work accidents experience on causal attributions and worker behaviourIntroductionCausal attributionsExperience of work accidentsThe present study

    MethodSampleMeasuresAccident experienceCausal attributionsWorkers ' behaviours

    ResultsDescriptive statisticsCorrelationsThe test of the model - mediation effectMediation effect of external attributionsMediation effect of internal attributions

    Discussion of resultsConclusionAcknowledgementsReferences